Shiplog Shiplog Orchestrator
Session orchestration, progress tracking, and background agent management
install
source · Clone the upstream repo
git clone https://github.com/aadivar/shiplog
Claude Code · Install into ~/.claude/skills/
T=$(mktemp -d) && git clone --depth=1 https://github.com/aadivar/shiplog "$T" && mkdir -p ~/.claude/skills && cp -r "$T/skills/shiplog" ~/.claude/skills/aadivar-shiplog-shiplog-orchestrator && rm -rf "$T"
manifest:
skills/shiplog/SKILL.mdsource content
Shiplog — Session Orchestration Skill
You have the Shiplog plugin active. This skill handles automatic project tracking, session orchestration, and background agent management.
Session Start Behavior
At the start of every session, you MUST:
- Read progress: Read
to understand the current project statedocs/PROGRESS.md - Display status: Show a one-line status summary:
Shiplog: Sprint {N} | Feature {ID} ({status}) | {done}/{total} features done - Load context: Use the information from PROGRESS.md to understand:
- What sprint we're on
- What feature was last worked on
- What was done in the last session
- What's planned next
CRITICAL: Post-Commit Agent Launching
After every
you perform, you MUST check for git commit
.shiplog/config.json. If it exists, launch background agents immediately. Do NOT wait for hook output. Do NOT skip this step.
How to launch agents after a commit
-
Read config: Read
to get agent model and which agents are enabled..shiplog/config.json- Handle both config formats:
- Nested:
{ "agents": { "model": "sonnet", "specs": { "enabled": true } } } - Flat:
{ "agentModel": "sonnet", "agents": { "specs": true } }
- Nested:
- Handle both config formats:
-
Read agent prompts: For each enabled agent, read its prompt file from the plugin's
directory:agents/agents/specs-agent.mdagents/prd-agent.mdagents/security-agent.mdagents/memory-agent.md
-
Launch agents using the
tool withAgent
andrun_in_background: true
:subagent_type: "shiplog:{agent}-agent"- Specs Agent → updates
docs/SPECS.md - PRD Agent → updates
anddocs/PRD.mddocs/PROGRESS.md - Security Agent → updates
docs/SECURITY.md - Memory Agent → updates memory files
- Specs Agent → updates
-
Launch in parallel: Launch specs, PRD, and security agents together. Memory agent can also run in parallel.
-
Don't block the user: Launch all agents in background and continue immediately. Never wait for agents to complete before responding.
Fallback: Hook trigger
If you see
SHIPLOG_AGENTS_TRIGGER output from a PostToolUse hook, treat it the same way — launch agents. But do NOT rely on this; always proactively launch after commits.
Custom Agents
If
.shiplog/agents/ directory exists in the project, also launch each custom agent found there.
Agent failure handling
- If any agent fails, it fails silently — do not interrupt the user's work
- Do not retry failed agents
- Do not report agent results unless the user asks
Available Commands
— Initialize Shiplog on current project/shiplog init
— Show project status dashboard/shiplog status
— Create a new feature from template/shiplog feature <name>
— Start a new sprint/shiplog sprint <n>
— Manually trigger all background agents/shiplog review
— Show/edit configuration/shiplog config
File Locations
- Config:
.shiplog/config.json - PRD:
docs/PRD.md - Specs:
docs/SPECS.md - Security:
docs/SECURITY.md - Progress:
docs/PROGRESS.md - Features:
docs/features/ - Agent prompts:
(in plugin root)agents/ - Templates:
(in plugin root)templates/